Safetensors Info (Readable):
SafetensorsInfoReadable lets you inspect the descriptive information stored in a Safetensors file without loading its model weights. It reads the file’s header to show basic file details, count tensor entries, and display metadata such as merge information or training settings when those details are present. This is useful when you want to identify or compare model files before using them in an image-generation workflow. The node presents metadata in a readable report and also makes it available as structured JSON. It can only show information actually stored in the file; it does not infer missing settings or verify that the model will work in a particular workflow.
Safetensors Info (Readable) Input Parameters:
file
Select a Safetensors file from the dropdown. The available options are discovered from the model folders configured for the node and include filenames ending in .safetensors or .sft; duplicate full paths are removed. This selection determines which file is inspected unless you provide a nonempty custom_path. If no files are found, the dropdown contains the placeholder (safetensorsが見つかりません) (“No Safetensors files found”), which is not a readable file. There is no numeric minimum or maximum and no fixed default file: the choices depend on the files available in your installation.
custom_path
Enter a path to a specific file when it is not available in the dropdown or when you want to inspect a different file. A nonempty value takes priority over file. The node removes surrounding whitespace and surrounding single or double quotation marks before using the path, which can help when pasting a quoted path. The default is an empty string, meaning the dropdown selection is used. No minimum or maximum length is specified. The path must point to an existing file that the application can read.
Safetensors Info (Readable) Output Parameters:
info
A readable text report showing the file’s name, path, size, number of tensor entries, and number of metadata items. It then lists each metadata key in sorted order with its value. If a metadata value is a string containing a valid JSON object or array, the report formats that value with indentation to make it easier to read. If the file has no metadata, the report says that __metadata__ is empty. This output is useful for reviewing file details directly or passing the text to another node that accepts a string.
json
An indented JSON string containing the file’s __metadata__ entries. Metadata values that are valid JSON objects or arrays stored as strings are converted into those JSON structures; other values remain as provided in the metadata. Unlike info, this output does not include the file path, file size, or tensor count. An empty metadata section produces {}. Use this output when a later step needs the metadata in a structured text format.
Safetensors Info (Readable) Usage Tips:
- Start with
filefor models already available in the dropdown. Leavecustom_pathempty so the selected file is the one inspected. - Use
custom_pathfor a file outside the listed model folders. Clear it when you want to return to the dropdown selection; otherwise, the custom path continues to take priority. - Compare the
tensorsandmetadatalines ininfowhen checking what a file contains. A tensor count can be present even when the metadata section is empty. - Use
infofor a quick human-readable review andjsonwhen another part of your workflow needs metadata as structured JSON text. Both are produced by reading the header, without loading the model weights. - If you edit a file and need to inspect it again, rerun the node. Its change check uses the file’s modification time.
Safetensors Info (Readable) Common Errors and Solutions:
モデルフォルダにsafetensorsがありません。custom_pathにフルパスを入れてください
- Explanation: The
filevalue is not a valid folder-and-filename selection, often because no Safetensors files were found and the dropdown contains only its placeholder. - Solution: Add a
.safetensorsor.sftfile to a configured model folder and select it, or enter the existing file’s path incustom_path.
ファイルが見つかりません: <path>
- Explanation: The chosen path does not point to an existing file. The file may have been moved, renamed, or selected from an out-of-date list.
- Solution: Check the filename and location, then select the file again or correct
custom_path. Confirm that the file is accessible from the machine running the node.
ファイルが小さすぎます
- Explanation: The file contains fewer than eight bytes, so the node cannot read the header length.
- Solution: Check that the file finished downloading or copying. Replace an incomplete file with a complete Safetensors file.
ヘッダサイズが不正です
- Explanation: The header length recorded at the start of the file is zero, larger than the file, or above the node’s 100 MiB header limit. This can indicate an incomplete or invalid file.
- Solution: Confirm that you selected the intended Safetensors file. If it may be damaged or incomplete, obtain a complete copy.
ヘッダがJSONとして読めません
- Explanation: The node could not decode and parse the header as JSON.
- Solution: Check that the file is a valid Safetensors file rather than a different file with a renamed extension. Replace it with a valid copy if necessary.
PermissionError or access denied
- Explanation: The file exists, but the application cannot open it to read its header.
- Solution: Check the file’s read permissions and ensure the application has access to its folder. You can also move a copy to an accessible model folder and select that copy.
